LLM functionality has improved significantly in recent quarters, enabling completely new interaction methods with software that was previously not possible. Today, we are proud to introduce an embedded ListAlpha Agent built on top of our MCP server architecture that allows our users to talk directly with their CRM about deals, adviser relationships and diligence notes.
We have designed the agent to be easily acessed from the start panel or by hitting the Ctrl + K shortcut. The launch dialog allows to reference any of the objects on the platform (deals, companies, etc.) which makes for a more robust experience over a standard ChatGPT or Claude text-baded LLM.
The Agent allows to offload the searching and analytical work typically done by a user to the LLM. For example, answering the question of “Which industry exec can advise us on this medical devices deal?” typically requires doing a number of searches in the contacts module and then carefully reviewing the results to choose the best contact.
Alternatively, the LLM agent is able to execute a number of parallel calls against the contacts module searching for “healthcare”, “medical devices” or “medtech” thus capturing a broader set of results.
In addition, the large context window allows the LLM to quickly read all of the career experiences of the returned executives and present the best answer in a structured manner (example below):
In order to make this feature even more powerful, we have introduced object embeddings both on the input and output of the LLM stream. When asking a question this allows to reference a particular deal, target company or an adviser, significantly narrowing down the scope of the search and making it more useful. Embeddings and citations on the output stream of the LLM allow users to interact with the model in a practical way, reviewing the results and even modifying retrieved records directly.
Our team continues to work on expanding the scope of the MCP server and the overarching Agent to include more objects and primitives from the platform.
One of the really promising directions we have identified is LLM-based analytics, which allows the LLM to directly query the database in order to create complex charts, such as time series analysis of deals split by sector. We have also begun work to integrate external research capabilities that allow to blend internal data with real world deal facts and public comps.
We tailor your onboarding experience to match what you're looking for. Tell us more about yourself and your team to help us out.
